Age and parity as determinants of maternal mortality — impact of their shifting distribution among parturients in Sweden from 1781 to 1980

Bulletin of the World Health Organization, 64 (1): 85-91 (1986) © World Health Organization 1986 Age and parity as determinants of maternal mortality impact of their shifting distribution among parturients in Sweden from 1781 to 1980 ULF HOGBERG' & STIG WALL2 The reduction in maternal mortality in Sweden between 1781 and 1980 is analysed with respect to changes in the distribution of age and parity among parturients over this period. Changes in maternal age contributed to almost 3% of the reduction in mortality over the period 1781-1911 and to 5% between 1911 and 1980. From 1965 to 1980, how- ever, about 50o of the reduction in mortality was caused by a decrease in maternal age. Changes in parity have had the opposite effect. Maternal deaths attributable to the risk factors of age and parity increasedfrom 46% during the 19th century to 80% from 1951 to 1980. In Sweden, age and parity are the major biological determinants of maternal mortality rate, with the optimum child-bearing age being 20-29 years. The risk of maternal mortality as a function of age can be represented by a J-shaped curve, with the risk of death for pregnant women younger than 20 years or older than 29 years beihg 2 to 6 times that of women in the optimum age group. Women at the upper age limit of fertility are more likely to develop com- plications, such as haemorrhage, toxaemia, or pro- longed labour, during pregnancy and parturition (1-3). The industrial countries have recently experienced a dramatic decline in maternal mortality rates, co- inciding with recent demographic changes -the drastic fall in birth rate and changes in the age and parity distributions of parturients. Previous studies have been inconclusive in their analysis of the effect of these changes on maternal mortality (4-8). The present study analyses the effect of changes in age and parity distributions of parturients on mortality rates and describes the role of age and parity as risk factors contributing towards maternal death. MATERIALS AND METHODS Since 1749, official statistics have been kept in Sweden on the distribution of births as a function of ' Department of Obstetrics and Gynaecology, University Hospital, S-901 85 Umea, Sweden. Requests for reprints should be sent to this author. 2 Associate Professor, Department of Social Medicine, Uni- versity Hospital, Umea, Sweden. maternal age (10-12). a Data on maternal mortality as a function of age are, however, less complete: during 1774-91 at 5-year age intervals,' 1911-20 at 10-year age intervals, and 1922-80 at 5-year age intervals (13). For the period 1973-80 the statistics were based on the number of births (12), but prior to this the number of confinements was used. This discrepancy is, however, unimportant for the analysis (14). Secular analysis of causes of death is associated with several sources of error. It was not until 1860 that the medical profession became responsible for issuing death certificates in Sweden; before then, the clergy had had the duty of registering the cause of death. The reliability of these statistics has been discussed previously (15-18). Indirect maternal deaths may have been included in the official statistics during the 18th and 19th centuries (19); however, deaths from abortions and ectopic pregnancies probably were not. Nevertheless, even in modern vital statistics indirect maternal deaths are sometimes erroneously reported as direct maternal deaths (20). In the secular analysis described here, no correction has therefore been made for indirect maternal deaths. The data available in Sweden on the distribution of births and maternal deaths by parity are less complete, and no national figures exist before the 1950s. However, the distribution of births by parity in seven Swedish parishes is available and the data cover 0.27% of all births in Sweden in the 19th a Tabellkommissionens Rikstabeller B II a: 1-12. 4640 85- U. HOGBERG & S. WALL century (20). These parishes were not selected as representative samples, but because they provide the only such data available on the situation in Sweden at that time. Nevertheless, the parishes are distributed geographically throughout Sweden and are situated in rural areas, where 90/o of the Swedish population lived in the 19th century. Since the distribution of births in the parishes as a function of maternal age does not differ from the national data, we assume that the national distribution of birth by parity follows the pattern in the parishes. National data on the distribution of births by parity cover 96% of all confinements for the period 1952-60, 75% for 1961-70, and almost 100%7o for 1973-80 (12, 21, 22). The distribution of maternal mortality as a function of age was obtained from the maternal death records in Sweden over the period 1951-80. Parity data are available in Sweden for 78% of maternal deaths for the periods 1951-60 and 1961-70, and for 84% for 1971-80. The missing information is assumed to be evenly distributed by age and parity. Since the number of deaths involved is small, the data have been divided into the following subgroups: mothers aged 15-29 and 30-49 years in combination with parity 1, 2-3, or >4. In order to estimate the impact of changes in age distribution on the maternal mortality rate, the direct method of standardization was used for the periods 1781-85 and 1911-15: the mortality rates for 1911-15 were used in combination with the birth distribution for 1781-85. Data for the 20th century were also standardized using the direct method. Adjusted mortality rates were calculated using the observed age-specific mortality rates for a given period with the age distribution of the common standard population, in this case 1911-15. The adjusted rates refer to the number of deaths that would have occurred in the standard population in this period had the mortality rate of a different period applied. In this way the age distribution remains fixed for the whole period of the study, and the use of standardized mortality rates avoids confounding from age. The difference be- tween the observed and adjusted number of deaths, calculated as described above, is attributable to the different age distributions. We define this difference as the number of "prevented" deaths. The propor- tion of maternal deaths prevented by the changes in age distribution is defined as the ratio of the pre- vented and adjusted number of deaths. The ratio of the adjusted mortality rate and the mortality rate of the standard population is referred to as the standard- ized rate ratio. Maternal deaths attributable to age and parity as risk factors are defined as all deaths in excess of the lowest risk found in the population (parity 2-3 and age 15-29 years). The etiological fraction of a certain exposure, e.g., parity and age, is expressed as (RR - 1)/RR, where RR is the risk relative to a reference category (23). RESULTS The agefactorfrom 1781 to 1911 The proportion of parturients aged 40-49 years decreased in Sweden from 11.707 to 8.5% from 1781 to 1911 (Fig. 1). The crude death rate per 100 000 births in Sweden over the period 1911-15 was 241.0. Given the distribution of births as a function of maternal age for the period 1781-85, we calculated the adjusted death rate as 247.3. The ratio of the ob- served number of deaths (1642) and the adjusted number (1691) is 0.971, i.e., 2.9%o of the reduction in mortality from 1781 to 1911 can be attributed to the change in age distribution of the parturients. The age factor during the 20th century In the 18th century, 43%o of births involved mothers aged 20-29 years. Today this proportion has increased to 65%7o, and the proportion of parturients ._ 0 8 -i C Ec 1911 1921 1931 1941 1951 1961 1976 -15 -25 -35 -45 -55 -65 -80 0 Fig. 1. Distribution of maternal mortality in Sweden at 5-year intervals from 1 91 1 to 1980 (crude death rates and age adjustment after standardization by the direct method). 86 AGE AND PARITY AS DETERMINANTS OF MATERNAL MORTALITY IN SWEDEN Table 1. Swedish maternal mortality from 191 1 to 1980 No. of maternal deaths/No. of births' 15-19 20-29 30-39 40 years years years years 53/4666 603/62093 731/57891 255/11643 83/4569 643/58950 701/49521 204/10865 64/5000 535/55110 584/44853 189/9541 85/4991 634/48808 670/36547 184/7531 60/4419 592/45165 614/32075 187/5713 50/5353 470/44324 541/34645 152/5288 36/6028 362/64593 455/46442 112/6085 20/7761 163/67757 241/45024 62/6136 14/8013 125/60620 148/36574 34/4663 8/9330 54/60657 87/33386 29/3783 7/13914 47/68999 40/29867 10/3183 3/11037 33/76386 34/19998 6/1881 1/8532 14/77464 15/22925 4/1133 1/5144 14/62780 16/27377 3/925 No. of maternal deaths 1642 1631 1372 1573 1453 1213 965 486 321 178 104 76 34 34 No. of births/ 5-year period 681465 619525 572520 489385 436860 448050 615740 633390 549350 535780 579815 546510 550270 481130 Mortality rate/ 100 000 births Crude Adjusted 241.0 241.0 263.3 264.2 239.6 241.7 321.4 327.5 332.6 347.3 270.7 169.3 156.7 169.8 76.7 86.4 58.4 66.8 33.2 43.9 17.9 23.3 13.9 24.0 6.17 13.2 7.1 12.7 Total 11082 11575 552 4.7 a The value shown for the number of births is the average annual value over each 5-year period. Table 2. Maternal mortality rates per 100 000 births by age and parity over the periods 1800-99, 1951-60, 1961 -70, and 1971-80 No. of deaths per 100 000 births Age (years) Parity 1800-99 1951-60 1961-70 1971-80 1 537 42.3 18.5 5.0 15-29 2-3 311 9.9 3.0 3.1 >,4 163 24.4 11.5 0.0 Crude rate 395 27.3 11.5 4.0 Parity adjusted 395 25.9 10.8 3.5 1 1160 170.5 84.7 22.5 30-49 2-3 636 52.3 18.2 13.1 ,> 4 630 72.8 15.5 14.0 Crude rate 678 78.8 25.2 14.0 Parity adjusted 678 75.7 22.2 6.2 1 Crude rate 694 63.3 25.2 7.2Age adjusted 694 74.8 35.3 9.6 2-3 Crude rate 482 27.7 8.2 6.3Age adjusted 482 32.1 11.0 8.3 Crude rate 585 61.4 14.9 3.9 Age adjusted 585 67.9 14.9 4.6 Overall crude rate 574 76.0 16.0 6.6 Age and parity adjusted 574 57.2 18.1 7.0 Period 1911 -15 1916-20 1921 -25 1926-30 1931 -35 1936-40 1941 -45 1946-50 1951 -55 1956-60 1961 -65 1966-70 1971-75 1976-80 No. of deaths 1642 1637 1384 1603 1517 1154 1046 547 367 235 135 131 90 87 No. of deaths "pre- vented" 0 6 12 30 64 81 61 46 57 31 55 56 53 Propor- tion of deaths IIpre- vented" (%) 0 0.37 0.87 1.87 4.2 7.7 11.2 12.6 24.3 23.0 42.1 62.2 60.9 87 U. HOGBERG & S. WALL AGE 10JO _ E 40 - 20 _ 10 _ 30 _ 0S1 O0 I PARITY 40 .- AGExPARITY I a 20 10 .2 a _____________ Changes in the maternal age distribution have there- fore been a major cause of the decline in maternal mortality since the 1940s in Sweden (Fig. 1). Parity and age factors during the 20th century The distribution of maternal deaths per 100 000 births as a function of age and parity over the period 1951-80 is shown in Table 2. The decline in maternal mortality over this period is slightly more pro- nounced for the primiparous and multiparous groups aged 30-49 years (cf. Fig. 3). 180 r 160 14444 1224 1 23.334 1 243-41 2.23 4 1 223 4 1 243 4 15-231 3-43 15-23n -4 15-23123-43 15-23 0-413 13w00-133 135-1060 1961-1070 1071-1330 Fig. 2. Distribution of maternal mortality by age (1 5-29 years, 30-49 years), parity (1, 2-3, >4), and age and parity over the periods 1800-99, 1951-60, 1961-70, and 1971-80. aged >,40 years has decreased to 1 o. The tendency of the majority of births to occur among women of optimum reproductive age is a phenomenon of the last 50 years, being especially marked over the last 20 years. Teen-age pregnancies experienced a transitory increase during the 1950s and 1960s, but in the 1970s, 89% of births occurred among women aged 20-34 years. To evaluate the effect on maternal mortality of the change in the distribution of births as a function of age during the 20th century, we used the direct method of standardization. Had the age distribution not changed, 11 575 maternal deaths would have been expected over the period 1911-80, in contrast to the 11 081 deaths observed. Thus 552 deaths were "'prevented" by a decrease in the average maternal age, i.e., a 4.70o decrease in the number of deaths (Table 1). The relative importance of the age- adjustment factor increased throughout the 20th century. Only 17% of the mortality decline caused by changes in maternal age occurred before the 1940s, but the age factor has been more pronounced since then. The proportion of maternal deaths prevented by the age-adjustment factor was 23%o in 1961-65 and 42%o in 1966-70, increasing to 67% in 1970-81. 1401 U$ .0 8 4- 8. 0 E cUC 0 +0 Age 15-29 *-----. Age 30-49 v - 120 _ 1001_ 80 _ 6oF 40 F 201_ 2-3 *1 1-..~ ~ ~ ~ ~ - -0- 1951-60 WHO 851345 Fig. 3. Trends in maternal mortality in Sweden by age (15-29 years, 30-49 years) and parity (1, 2-3, >4) from 1951 to 1980. 1961-70 1971-80 88 AGE AND PARITY AS DETERMINANTS OF MATERNAL MORTALITY IN SWEDEN 1800-99 1951-80 30- 49 yers 15 - 29 years \ 30-49years /4 I -, ._ - 1 2-3 4- 1 2-3 4- Parity Fig. 4. Comparison of age and parity as risk factors of maternal mortality in Sweden over the periods 1800-99 and 1951-80 (relative risk 1 = age 15-29 years and parity 2-3). If the maternal age distribution for 1800-99 is taken as the standard population, the fraction of deaths prevented by the age factor is 17.8% for the period 1951-80, 4% less than the value calculated using the age distribution for 1911-15 as the standard population (Table 1). Important changes in the distribution of births by parity occurred in the 20th century in Sweden. The proportion of primiparous pregnancies increased from 237o in 1899 to about 450o in 1980, but the proportion of pregnancies among women aged > 30 years remained approximately constant at 5.7%; however, the proportion of multiparous pregnancies (,>4 deliveries) decreased from 43% in 1899 to 5% in 1980 (Fig. 2). The effect of the changes in the distribution of births by parity on maternal mortality is dichotomous: on the one hand there is a tendency towards fewer multiparous pregnancies, but, on the other, there is a tendency towards an increasing proportion of primiparous pregnancies among women aged 15-29 years, with a constant proportion of primaparous pregnancies among women aged 30-49 years. However, the effect of the primiparous pregnancies outweighs that of the decreasing number of multiparous pregnancies (Table 2, Fig. 2 and 3). The adjusted number of deaths calculated for the period 1951-80 using the parity distribution for 1800-99 was 52 (etiological fraction 7.0O1o), resulting from a reduction in the proportion of multiparous parturients. The proportion of maternal deaths prevented between 1951 and 1980 by the combined effects of the changes in age and parity distributions was 16.7%, which is almost the same proportion prevented by changes in age alone (17.8%). The proportion of deaths prevented by the combined effect was especially pronounced in the period 1951-60, during which 66% of the total deaths prevented from 1951 to 1980 occurred. The changes in the age and parity distributions operate inter- actively since 150 deaths were prevented by the change in the combined distribution, while only 110 were prevented by their separate effects. Age and parity as risk factors By identifying the group exposed to the least risk of death during the 19th century (age 15-29 years, parity 2-3), maternal deaths caused by the increasing risk of age and parity can be calculated. In this way the etiological fraction of maternal mortality was 46% in Sweden during the 19th century. This fraction increased to 77% of maternal deaths over the period 1951-80. In other words, the relative importance of age and parity is greater today than in the 19th century. The risk difference between the age groups 15-29 and 30-49 years was 3-fold during the 19th century, but 21-fold during 1951-80 (Fig. 4). DISCUSSION Changes in the death-cause registration over the period 1780-1980 in Sweden may have had a bearing on the results reported here. However, any bias resulting from this should have been more pro- nounced during the 18th and 19th centuries when the impact of the age-factor analysis of the official statistics was lowest. Furthermore, the analysis of the age factor from the seven parishes gave results that were comparable to the official statistics in the 19th century. This indicates that the secular parity changes were also not biased. Parity For the assessment of obstetric management and evaluation of medical techniques there is a need for studies of the effect of biological determinants on the outcome of pregnancy (24). It has been suggested that multiparity is associated with a higher level of risk than primiparity. Our results indicate that shifts in the parity distribution of parturients have had a negative influence on the decline of maternal mortality. However, the combined effect produced by changes in maternal age and parity counteracts that of changes in parity alone. Age The risk of maternal mortality as a function of age can be represented graphically by a J-shaped curve, 89 .X0 0 a 'a w U. HOGBERG & S. WALL with a higher risk of mortality at the extreme ages of fertility. The trough of the curve is shallower, i.e., less pronounced, in countries with high birth rates. One possible explanation for this may be that in countries with low birth rates the parturients are predominantly in the age group 20-34 years, whereas older women belong to a high-risk group that is susceptible to intercurrent diseases and has less access to health services (26). Maternal mortality is there- fore increased by parturients of the socioeconomic high-risk group who are in the upper reproductive age category (27, 28). Data from Sweden are consistent with a trend towards an increase in the relative mortality risk by age at a time when total mortality in the country is declining. Swedish data also deviate from the classic J-shaped curve, since teenagers benefit more than other age groups from medical services (25). This also confirms earlier suggestions that the effect of increasing age is the most difficult risk factor to minimize (9). Evaluations of the epidemiological effect on maternal mortality of the shift towards younger parturients is more difficult. In several countries national or regional data have been analysed in an attempt to estimate the impact of birth distributions by age and parity on maternal mortality. For example, in Australia no relationship was found between birth rate and maternal mortality in the early part of the 20th century (5). However, in the USA changes in the age and parity distribution have had an influence on maternal mortality over the period 1919-69 (6). In Scotland changes in maternal age and parity over the period 1939-46 had no effect on maternal mortality (8). Finally, in the Netherlands, 7% of the total decline in maternal mortality from 1950 to 1965 can be attributed to a shift towards younger mothers (7). Our results show that 27%7o of the reduction in maternal mortality rate over the period 1951-80 in Sweden can be attributed to changes in the age distribution of the parturients. These results are comparable to those obtained from studies on infant and perinatal mortality, which indicate that 25-50% of the reduction in mortality is caused by a shift towards younger mothers and fewer births per mother (27,28). However, in Sweden only 9% of the reduction in perinatal mortality during 1953-75 can be attributed to demographic influences (14). In conclusion, our results indicate that before the 20th century changes in maternal age played only a small part in reducing maternal mortality, but that since the 1930s the effect of maternal age has become increasingly significant: over the last 15 years as much as 50% of the reduction in maternal mortality in Sweden can be attributed to this factor. In contrast, changes in the parity distribution have had the opposite effect; however, the combined effect of maternal age and parity counteracts that of parity alone. ACKNOWLEDGEMENTS This work was supported by grants from the Swedish Agency for Research Cooperation with Developing Countries, by the Foundation Allmanna BB:s Minnesfond, and by the Foundation J. C. Kempes Minne. RESUME AGE ET PARITt EN TANT QUE FACTEURS DETERMINANTS DE LA MORTALITt MATERNELLE-IMPACT DE L'EVOLUTION DE LEUR DISTRIBUTION CHEZ LES PARTURIENTES EN SUEDE, DE 1781 A 1980 Entre 1781 et 1980, le taux de mortalite maternelle a baisse de facon spectaculaire en Suede, coincidant avec une re- duction du nombre moyen d'enfants mis au monde par femme, qui est passe de 4,9 a 1,4. L'article analyse l'effet des modifications de la distribution des parturientes par age et parite sur le taux de mortalite et d&crit les tendances seculaires des distributions de l'age et de la parite en tant que facteurs de risque contribuant aux deces lies a la maternite. On ne dispose pas de donnees sur les deces maternels en Suede en fonction de l'age et de la parite pour l'ensemble de la p6riode d'etude. Par exemple, tous les deces maternels ont ete enregistres entre 1951 et 1980, mais on ne dispose que de donnees partielles sur le nombre total de naissances au 19eme siecle. On a examine les eventuels biais pouvant fausser l'analyse des donnees. L'impact des modifications de la distribution par age sur le taux de mortalite maternelle a e calcule au moyen de la methode de standardisation directe, qui a egalement e utilisee pour examiner l'effet 90 AGE AND PARITY AS DETERMINANTS OF MATERNAL MORTALITY IN SWEDEN 91 conjoint de I'age et de la parite sur ce taux. De 1781 a 1980, la proportion de parturientes de 35 ans et plus est passee de 32% a 7%. Entre 1781 et 1911, les modifications de la distribution des parturientes par age pouvaient expliquer presque 3% de la baisse du taux de mortalite, et 5 01 entre 1911 et 1980. Toutefois, entre 1965 et 1980, environ 50% de la reduction du taux de mortalite maternelle etait due a une diminution de l'age maternel moyen. Au cours des deux siecles de la periode d'etude, la proportion des multipares (femmes ayant donne naissance a 4 enfants ou plus) est passee de 43% a 5%, alors que la pro- portion de primipares s'est elevee de 23% a 45%. Les modi- fications de la parite au cours de la periode d'etude ont eu pour effet d'augmenter le taux de mortalite mais, globale- ment, les modifications des distributions par age et selon la parite ont conduit a une reduction de la mortalite mater- nelle. La proportion des deces maternels attribuables aux facteurs de risque age et parite est passee de 46% au 19eme siecle a 80% au cours des 30 derni&es annees. REFERENCES 1. NORTMAN, D. Parental age as a factor in pregnancy outcome and child development. Reports on population/family planning, 16: 1-52 (1974). 2. EASTMAN, N. J. The hazards of pregnancy and labor in the "grand multipara". New York journal of medicine, 40: 1708-1712 (1940). 3. OXORN, H. Hazards of grand multiparity. Obstetrics and gynecology, 5: 150-156 (1955). 4. YERUSHALMY, J. Studies in child birth mortality-age and parity as factors in puerperal fatality. Public health reports, 55: 1195-1220 (1940). 5. DUBLIN, M. Maternal mortality and the decline of the birth rate. Annals of the American Academy of Political and Social Science, 188: 107-116 (1936). 6. BERRY, L. G. Age and parity influences on maternal mortality: United States, 1919-1969. Demography, 14: 297-310 (1947). 7. BONTE, J. T. P. & VERBRUGGE, H. P. Maternal mortality-an epidemiological approach. Acta obste- tricia et gynecologica Scandinavica, 46: 445-474 (1967). 8. McKINLEY, P. L. The influence of changes in the age and parity constitution of mothers on the trend of foetal and maternal mortality in recent years. Health bulletin, 8: 16-19 (1950). 9 AUGENSEN, K. & BERSJO, P. Maternal mortality in the Nordic countries. Acta obstetricia et gynecologica Scandinavica, 63: 115-121 (1984). 10. Bidrag till Sveriges Officiella Statistik A/ Befolknings- statistik, 1861-1910. 11. Vital statistics, Swedish Official Statistics, 1911-1972. Stockholm, Central Bureau of Statistics. 12. Medical Birth Registration, Statistical Reports, 1973-1980. 13. Causes of death, Swedish Official Statistics. Stock- holm, Central Bureau of Statistics 1911-1980. 14. MEIRIK, 0. ET AL. Impact of changing age and parity distributions of mothers on perinatal mortality in Sweden, 1953-1975. International journal of epi- demiology, 8: 361-364 (1979). 15. DANELL, C. The demographic data base at Umea University. In: Brandstrom, A. & Sundin, J., ed., Tradition and transition, studies in microdemography and social change, UmeA, 1981. 16. Official Statistics of Sweden. Public health in Sweden 1954-1970. Stockholm, National Board of Health and Welfare. 17. Soop, E. Birth by parity. Statistisk Tidskrift, 4: 640-645 (1955). 18. LEVIN, M. L. The occurrence of lung cancer in man. Acta unio internationalis contra cancrum, 9: 531-541 (1953). 19. FRAZER, C. Selected perinatal procedures. Acta obstetricia et gynecologica Scandinavica, Suppl. 117, 1983. 20. BUCHANAN, R. Effects of childbearing on maternal health. Population reports, 8: 125-139 (1975). 21. TIETZE, C. Maternal mortality. World health statistics quarterly, 30: 312-339 (1977). 22. SHAPIRO, S. ET AL. Infant, perinatal, maternal and childhood mortality in the USA. Boston, Harvard University Press, 1968. 23. LLEWYLLYN-JONES, D. Human reproduction. London, Faber & Faber, 1974. 24. SCHAFFNER, W. ET AL. Maternal mortality in Michigan: an epidemiologic analysis, 1950-1971. American journal ofpublic health, 67: 821-829 (1977). 25. BERNHARD, R. P. ET AL., Risk of maternal death in Indonesia and Sweden: patterns and trends in 3D- display for teaching. Sozial- und Praventivmedizin, 29: 172-173 (1984). 26. WEBB, J. & WESTON, E. P. Recent trends in maternal mortality. Medical officer, 1: 201-202 (1951). 27. MORRIS, N. M. ET AL. Shifting age-parity distribution of births and the decrease in infant mortality. American journal ofpublic health, 65: 359-362 (1975). 28. GENDELL, M. & HELLEGERS, A. E. The influence of the changes in maternal age, birth order and color on the changing perinatal mortality, Baltimore 1961-1966. Health services reports, 88: 733-742 (1973).